From Cognitive Science

For the most part, my interest in Retrieval Practice and Successive Relearning has been from a Cognitive Science, Behavioural Psychopathology perspective, on retrieval, re‐encoding and reconsolidation.

The New Theory of Disuse (NTD) (Bjork & Bjork, 2011) discusses memory through two strength properties: retrieval strength and storage strength. Retrieval strength refers to how readily an item can be accessed in memory at the moment, whereas storage strength refers to how well learned that item is in memory. Retrieval practice enhances both retrieval strength and storage strength, but the degree to which these strengths are enhanced depends on the level of each at the time of the test. For instance, gains in storage strength are largest when retrieval strength is low (which corresponds to more retrieval effort). Conversely, the higher the storage strength, the smaller the gains in retrieval strength. Once accumulated, storage strength is never lost. Retrieval strength fades over time, but the extent to which it fades depends upon the level of storage strength. A higher level of storage strength minimises losses in retrieval strength; however, a low level of storage strength means that any gains in retrieval strength will be short-lived without additional learning. In sum, NTD posits that retrieval strength and storage strength dynamically influence the benefit of a learning trial.

Elaborative retrieval accounts posit that the retrieval of information from memory results in memory elaborations that provide new routes to access the information, and that these additional retrieval routes make future attempts to access the information more likely to be successful (Carpenter and DeLosh 2006; Carpenter 2009). Here, the semantic information is the key mechanism by whichtesting promotes learning. Critically, these mediated pathways enhance the likelihood of future recall because any of the mediating information can now serve as a potential route to the target.

Episodic context accounts (as highlighted by Karpicke) posit the process of retrieval serves to increase the specificity of the search process by adding unique contextual elements to the trace thereby making retrieval easier on subsequent tests.

Other theories high the re‐encoding process postulated to occur after a successful retrieval attempt.

Reconsolidation accounts (Finn and Roediger 2011) posit that when information is retrieved from memory, it enters a labile state, rendering it amenable to change (Dudai 2004). That after the first successful recall of the studied information, the retrieved information enters an unstable state, thereby enabling the memory trace to be strengthened by the postretrieval re‐encoding of the correctly retrieved information, (Eriksson et al., 2011).

Theoretical models and a vast body of research long emphasise the role of retrieval and in memory updating, more recently the neural mechanisms and basis behind these process or processes are currently being explored by Neuroscience researchers using neuroimaging evidence Functional Magnetic Resonance Imaging (fMRI) and event-related potentials (ERPs).

To Cognitive Neuroscience

As succinctly as possible: Brain activation patterns during the first test or recall of "studied pairs predicts performance on the second test." Importantly, while subsequent memory analyses of encoding trials also predict later accuracy, the brain regions involved in predicting later memory success are "more extensive for activity during retrieval (testing) than during encoding (study). Those additional regions that predict subsequent memory based on their activation at test but not at encoding may be key to understanding the basis of the testing effect (Liu et al., 2014).

Regions in prefrontal and parietal cortex and medial temporal lobe (MTL) consistently emerged Liu et., (2014) analyses and these regions have been associated with conceptual and attentional processes and memory storage, all of which are required for successful learning. Specifically, researchers were observing the differences between subsequent memory effects for encoding and retrieval subsequent memory effects to explain why testing is superior to study.

Liu et al., (2014) also prompted our awareness of response time and the acceleration of correct recalls with subsequent correct retrievals. An observation we had also noticed in field settings.

Correct recalls in Test 2 were significantly faster than at Test 1, (in part greater task familiarity with the task interface) though not accelerated when correct at Test 2 vs. incorrect at Test 2.

Predefined fMRI analysis: Could brain activity during retrieval (testing) predict subsequent memory performance?

Liu et al., (2014) reported that while being correct on the first cued recall test did not guarantee correct recall on the second test, the activation values in brain activation on the first successful recall did predict whether the second attempt would also be correct. (Of six brain regions measured significant correlations were observed in right Prefrontal Cortex (PFC) (r = .64, p = .022) and right Posterior parietal cortex (PPC) (r = .57, p = .012) Activations in the other 4 regions were also positively correlated with behavioral performance on Test 2 although not statistically reliable (left PFC: r = .37, left PPC: r = .42, left hippocampus: r = .17, right hippocampus: r = .11). Brain regions that were more active during testing than during restudying.

Furthermore Wing et al. (2013) and van den Broek et al. (2013), regardless of the modified recall method, reported that same brain regions emerged during successful retrieval (on the intermediate tests). In the van den Broek et al. (2013), significantly better and faster for tested words than for restudied words. Conceivably, the activation of right PFC during retrieval is responsible for stronger association formation and better learning than what typically occurs during study.

van den Broek et al. (2013), identified enhanced activity in inferior frontal gyrus (IFG) during testing than restudying, repeatedly related to intentional, non-automatic processing in memory studies. Higher activation during testing than during restudying therefore supports the idea that testing involves more intentional, effortful processing than restudying, but as such, was not predictive of better memory retention.

Does learning during testing differ from learning during study?

Significant subsequent memory effects were found in left Prefrontal Cortex and bilateral hippocampus, left, t(14) = 2.66, p = .01; right, t(14) = 2.12, p = .027.There were no significant effects in right PFC or right PPC, t's <1.

In summary:

Both the Regions of Interest (ROI) and the whole brain exploratory analysis revealed that the brain regions previously identified as responsible for learning during study, namely the left PFC, left PPC and hippocampus were also identified as regions responsible for successful encoding. Importantly, these regions were also involved during the testing phase, suggesting that participants could also learn from testing without feedback and re-study. Furthermore, Liu et al., (2014) identified additional brain regions that are only activated during retrieval yet also predict subsequent correct recall.

Whilst I am currently seeking expert contribution to this section, here is a first summary.

Retrieval practice is an effortful process as compared to simple restudy. Consistently, neuroimaging studies have found greater neural activity during retrieval than during restudy (Wing et al., 2013). In addition, retrieval practice could also potentiate subsequent learning, which is associated with greater frontoparietal activity (Nelson et al., 2013). These results suggest that during retrieval practice, the co-activation of old and new memories might provide a unique opportunity to modify these representations and to facilitate memory updating.

Researchers have proposed a two-process account (Liu and Reder, 2016; Liu et al., 2018) for the testing effect that we look to exploit with learner outcomes. Learners undergo a re-encoding process through re-exposure to the correct answer after successful retrievals (post-retrieval re-encoding), in addition to the retrieval attempt process. This post-retrieval re-encoding process also involves metacognitive monitoring and self-evaluation (Bai et al., 2015; Zhang et al., 2018) and can further enhance testing-related benefits (Liu and Reder, 2016).

Ye et al., (2020) confirm this conclusion. It was described as an "elegant human study" Ye et al., (2020) examined the effects of retrieval practice on memory performance and neural responses. The results from a set of experiments show that retrieval practice strengthens new memories and reduces intrusions of old memories without suppressing the old memories. Interestingly, this was related to enhanced representations in medial prefrontal cortex, further supporting the idea that this region of the brain is important for memory integration and consolidation.

Ye et al., (2020) reported retrieval practice as superior for long-lasting memory, recalling more updated targets with shorter response times and showed fewer memory intrusions when compared to Restudy. Distinct patterns and brain regions for the Retrieval Practice and Restudy were identified. Correctly recalled items showed stronger target activation than did incorrectly recalled ones in the ventral temporal cortex (VTC) during restudy whereas retrieval practiced items that were later correctly recalled were associated with stronger target activation than incorrectly recalled ones in the medial prefrontal cortex (MPFC).

This finding suggests that successful memory updating may involve different representations under the Retrieval Practice and Restudy conditions. The greater association of MPFC representation with enduring retention is consistent with consolidation processes and suggests that retrieval practice improves updating by driving consolidation more successfully than does restudy.

Liu et al., (2014) fMRI studies showed that testing compared with re-study, involved more monitoring and working memory-related brain activity. Findings suggest that learners may benefit from practice tests prior to an exam, which not only improve their exam performance, but also allow for better metacognitive monitoring based on their subjective experience during the practice test.

Regions associated with retrieval attempts were found to always predict subsequent memory success (the greater the activation, the more likely the correct recall); however, the regions associated with re-encoding would sometimes predict subsequent failure, specifically when subjects had correctly recalled the associated word several times already.

Their results suggest that whether a testing effect advantage is observed depends on both on the retrieval process and the re-encoding process which follows that retrieval. Based on this pattern, these researchers proposed two distinct processes underlay the testing effect, namely a retrieval process that brings the answer to mind and a re-encoding process for additional study after retrieval.

Feng et al., (2019) suggest that spaced learning (successive relearning) improves long-term memory by increasing retrieval effort and enhancing the "pattern reinstatement of prior neural representations," with extended repetition lags helping to eliminate the residual representation in working memory. Supporting the study-phase retrieval hypothesis and emphasising the role of neural pattern reinstatement in strengthening memory via repeated study.

Antony et al (2017) propose that retrieval acts as a fast route to memory consolidation. Specifically, that retrieval integrates the memory with stored neocortical knowledge and differentiates it from competing memories, thereby making the memory less hippocampus dependent and more readily accessible. That retrieval rapidly consolidates memories by coactivating related knowledge structures to facilitate hippocampal-neocortical representations via the ‘online’ reactivation of associative information, similar to those processes that occur during sleep and offline consolidation.

The nature of learning driven by coactivation depends on how strongly memories are activated: strong coactivation of memories leads to integration of those memories, whereas moderate activation of competing memories triggers their adaptive weakening and pushes retrieved and competing memories apart in representational space leaving the retrieved memory in a distinct, accessible state for future recall. Importantly, restudy (i.e., simple re-exposure to a complete, previously stored memory) does not share these computational characteristics. Restudy may re-impose the memory’s original pattern onto the hippocampus and neocortex, causing some strengthening of the original trace. However, because restudy triggers less coactivation of related memories it does not adaptively shape the hippocampal–neocortical memory landscape in the same way as active retrieval.